| 摘 要: 针对帕金森病(PD)严重程度评估依赖医生主观判断的局限性,本文设计并实现了一种客观的震颤监测与评级系统。该系统主要包括可穿戴设备、App、云平台及评估模型4个部分。可穿戴设备负责采集震颤数据,并将数据传输至App和云平台进行存储,同时支持模型训练。系统采用带限多重加权傅立叶线性组合算法(BMWFLC)对数据进行预处理,并提取时域和频域共计26个特征。随后,通过机器学习对数据进行多分类分析。实验结果显示,该系统达到了97.0%的加权精度、94.0%的加权召回率、93.9%的加权灵敏度、99.6%的加权特异度及94.5%的加权F1值。以上结果表明,该系统能够客观且准确地反映患者帕金森震颤的严重程度。 |
| 关键词: 帕金森震颤;物联网;BMWFLC;希尔伯特黄变化;机器学习 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 浙江省软科学研究计划项目(2024C25058);浙江省自然科学基金探索公益项目(LTGY23H170004) |
|
| BMWFLC-based Parkinson's Tremor Monitoring and Rating System |
|
ZHANG Huaqing1, YUE Xingyu2, JIANG Lurong2, QI Pengjia2, YANG Jianbin2, TONG Jijun2
|
(1.Department of Clinical Medical Engineering, The Second Afiliated Hospital of Zhejiang University School of Medicine, Hangzhou 310009, China;
2.School of Information Science and Engineering, Zhejiang Sc-i Tech University, Hangzhou 310018, China)
zhqnature@zju.edu.cn; yxy@luxuankj.com; jianglurong@zstu.edu.cn; qipengjia@zstu.edu.cn; 202130504193@mails.zstu.edu.cn; jijuntong@zstu.edu.cn
|
| Abstract: To address the limitations of subjective physician-dependent assessments in evaluating Parkinson's disease (PD) severity, an objective tremor monitoring and rating system was designed and implemented. The system primarily consists of four components: a wearable device, a mobile application (App), a cloud platform, and an evaluation model. The wearable device is responsible for collecting tremor data, transmitting it to the App and cloud platform for storage, and supporting model training. The system employs the Band-limited Multiple Weighted Fourier Linear Combiner (BMWFLC) algorithm to preprocess the data and extracts a total of 26 time-domain and frequencydomain features. Subsequently, machine learning is applied for multi-class classification analysis of the data. Experimental results demonstrate that the system achieves a weighted accuracy of 97.0%, a weighted recall of 94.0%, a weighted sensitivity of 93.9%, a weighted specificity of 99.6%, and a weighted F1 score of 94.5%. These results indicate that the system can objectively and accurately reflect the severity of Parkinson's tremor in patients. |
| Keywords: Parkinson's tremor; Internet of Things ( IoT); BMWFLC; Hilbert-Huang Transform ( HHT);machine learning |
